A survey of support for structured communication in concurrency control models

• We survey communication among execution flows in concurrency control models.
• We evaluate each model’s support for structured communication.
• Most models do not guarantee that the communication among execution flows occurs only in syntactically restricted code regions.
• The ease of reasoning about communication among execution flows is often neglected.
• Structured communication could reduce the complexity of concurrent programming.

The two standard models used for communication in concurrent programs, shared memory and message passing, have been the focus of much debate for a long time. Still, we believe the main issue at stake should not be the choice between these models, but rather how to ensure that communication is structured, i.e., it occurs only in syntactically restricted code regions. In this survey, we explore concurrency control models and evaluate how their characteristics contribute positively or negatively to the support for structured communication. We focus the evaluation on three properties: reasonability, which is the main property we are interested in and determines how easily programmers can reason about a concurrent program’s execution; performance, which determines whether there are any distinct features which can prevent or facilitate efficient implementations; and composability, which determines whether a model offers constructs that can be used as building blocks for coarser-grained, or higher-level, concurrency abstractions.